Towards improved automatic chemical kinetic model reduction regarding ignition delays and flame speeds

In chemical kinetic model reduction under internal combustion engine conditions, most implementations only consider ignition related chemistry without consideration of flame speed prediction. In practice, flame propagation commonly exists in spark ignition engines, dual-fuel with pilot injection compression ignition engines, reactivity controlled compression ignition engines, and etc. Due to the inherent time-consuming nature, it is impractical to run a 1-D flame code with trial-and-error methods for model reduction, especially when starting with a large chemical kinetic model. In this paper, an improved reduction methodology is proposed for construction of a small set of species that give accurate predictions of both flame speeds and ignition delays. First, a strong correlation is found between the errors of maximum H radical and the errors in prediction of laminar flame speeds. Addition of H to the search targets in graph-based methods is conducted showing improvement in accuracy of flame speed prediction. Second, the normalized flame speed sensitivity with rate constants is analyzed for identifying a set of species that strongly influences the prediction of flame speeds. Finally, a trial-and-error based method is used for further reduction with a 0-D testbed for prediction of ignition only, while keeping the species important to flame chemistry. The newly proposed reduction methodology is used for development of accurate skeletal models predicting both ignition and flame speeds for several hydrocarbon fuels. These skeletal models include methane (27 species), propane (32 species), n-heptane (126 species), and primary reference fuel gasoline surrogates (207 species) with high fidelity to be used in engine simulations.